license with patent grants appropriate for specifications?

[Sam Hartman]

>>>>> "Ian" == Ian Jackson <ijackson at chiark.greenend.org.uk> writes:

    Ian> Lawrence Rosen writes ("RE: license with patent grants
    Ian> appropriate for specifications?"):
    >> The compromise I try to reach with patent owners in the context
    >> of industry standards is that (1) patent owners may choose to
    >> limit the scope of their patent license grants to compliant
    >> implementations, and (2) those patent licenses may not restrict
    >> anyone from creating and distributing non-compliant
    >> implementations.

    Ian> This is disingenuous.

    Ian> The only way to read those two statements so as not to be
    Ian> directly contradictory is to infer that point 2 means that
    Ian> the patent licences fail to licence non-compliant
    Ian> implementations but do not contain an explicit contractual
    Ian> term forbidding the licensee from producing non-compliant
    Ian> implementations.  Of course, such a term would (depending on
    Ian> its terms) very likely be heinously awful.

    Ian> But what your point 2 _appears_ to say is that the patent
    Ian> fails to restrict non-derivative works, which is a direct
    Ian> contradiction of point 1.

    Ian> So I say that your posting is misleading.

    >> Point 2 is necessary so as to be consistent with open source
    >> freedom to create derivative works. Point 1 recognizes that
    >> patent owners and standards organizations are usually not
    >> interested in expressly granting broad patent licenses for
    >> non-compliant implementations.

    Ian> These two points are in fact in direct opposition.  The
    Ian> freedom to create and distribute noncompliant derivative
    Ian> works is one of the freedoms that define Open Source (or
    Ian> Free) software.  If there is a patent that prevents this then
    Ian> the software is not Open Source.

It's a bit more complicated than that.  There are a lot of folks in
the open-source community that ignore patents to one degree or
another.  I'm most familiar with Debian's policies in this matter, but
I know other projects have similar requirements.

What Debian seems to do is isomorphic to the following.  Debian looks
at a license assuming that there are no patents on the software and
considers whether the license is free.  In this analysis a patent
grant applicable only to compliant implementations [*] would be OK, but a
license condition requiring you only produce a compliant
implementation would not be OK.  I.E. the license can say you get
these open-source copyright licenses and in addition if your
implementation happens to be compliant you get these other patent
licenses.  A license that required your implementation be compliant
and gave you both the copyright and patent licenses would be
problematic.

Patents are of course not completely ignored.  If it looks like a
particular patent is going to be enforced, the analysis gets more
complicated.  There are a lot of patents out there.  You cannot assume
all of them are valid.  Moreover most open-source projects do not have
the resources to determine whether some software actually infringes
some particular patent.  Putting it mildly, there is significant risk
in the strategies people adopt in practice.



[*]  The case that has actually been analyzed is a patent grant for
patents infringed by the software as originally shipped.  Some in the
Debian community would probably argue that the compliant
implementation requirement is somehow more odious.
In this respect, Debian, like the IETF is consensus based.  As we are
aware, building consensus on IPR issues is challenging.

--Sam